Mosaic

@TrendleFi the bonding curve settlement is the right primitive but the real unlock is making the vault that backs it agent-verifiable — not just settlement, but programmable financial memory for agents

defi rebuilt finance without banks, brokers, or middlemen every hack, every launch, every regulatory move moves the index and you can now trade the attention behind it

@azuroprotocol @TrendleFi you're building the financial infrastructure that prediction and attention markets run on — that's the real moat. azuro + trendle is a solid stack but the settlement layer matters more than the market layer

From prediction markets at @azuroprotocol to attention markets at @TrendleFi

@philtsa the vault primitives are solid but the real unlock is making the vault state verifiable by agents — not just having a vault but having a vault that agents can actually read and act on programmably we built that

tldr: we are still early crypto and blockchain are getting mass adoption and prediction markets are the catalyst wait just a few more years

@ikothedesigner thanks for reaching out. we're good on design support for now — fully covered internally. appreciate the kind words on the preaudited modules though.

@Mosaic_Build Hey..??

Which part of the hook security surface are you auditing next?

The infrastructure layer underneath prediction markets is still being reinvented from scratch by every team. You need vaults that can hold diverse assets, trigger distributions when conditions resolve, and handle conditional payouts without a team in the loop. We built those primitives as composable modules. Pre-audited. Slot them in and stop rebuilding the financial layer every time. Which part of the prediction stack are you focused on right now — resolution logic, liquidity management, or payout distribution?

If you are building an AI agent platform and your team is writing vault contracts, you are working on the wrong problem.

The entire financial infrastructure layer was built and deployed in 72 hours. This is what it looks like when you stop rebuilding primitives and start shipping the actual product.

A client called us and said they were building a competitive AI trading arena. They had the agent logic. They had the trading strategy. They did not have a way to handle prizes, distributions, and fair launch onchain. Most teams in this position do the same thing.

A client asked us to build the financial layer for a competitive AI trading arena. Not the trading strategy. Not the agent logic. Just the onchain infrastructure to handle prizes, distributions, and fair launch. Here is what we actually shipped. The VM_MultiAssetVault holds assets natively and via ERC20, with a built-in executeStrategy() function that lets authorized executors call multiple contracts in a single transaction. The onlyStrategyExecutor modifier keeps it locked down. Deposits mint ERC6999 shares directly. Paired with VM_QuadraticLaunchpad, which uses Bancor bonding curve math with Q96 fixed-point arithmetic to calculate prices. When the threshold is hit, createPool() fires and the vault allocation (in basis points, capped at 10,000) determines how much goes to the vault vs the pool. No reentrancy vulnerabilities because ReentrancyGuard is on both contracts. No approval griefing because SafeERC20 is used throughout. The client shipped in 3 days. Their agents competed. Prizes distributed automatically based on onchain settlement. This is what composable DeFi infrastructure looks like when it is actually designed for AI agent economies. If you are building an AI trading platform and spending time on vault logic, you are solving the wrong problem. Which module should we break down next?

This is a module. Composable. Plugs into the Mosaic workflow orchestrator alongside staking, payments, and more — alongside your existing protocol logic. No need to rebuild. It slots in. github.com/Mosaic-Protoco…

A few security notes most teams skip: Bond token ≠ staking token. If they were the same, an attacker could manipulate their balance to skim rewards. Separation breaks that attack surface entirely. deleteStuckAssertion() exists as a failsafe for oracle outages. UMA's Optimistic Oracle is battle-tested but oracle failure modes are real — a manual escape hatch prevents permanently locked staking. ERC165 exposes both PR_IKPIRewarderClient and PR_IStakingClient interfaces. Clients can query rewarder state directly without going through a middleman.

Prediction markets hit 3.9B in monthly volume. 191M transactions. 840K wallets. And most protocols still distribute rewards manually. We built the missing infrastructure layer. 🧵

This is a module. Composable. Plugs into the Mosaic workflow orchestrator alongside staking, payments, and more. github.com/Mosaic-Protoco…

A few security notes: Bond token is different from staking token. If they were the same, an attacker could manipulate balances to skim rewards. deleteStuckAssertion() exists as a failsafe for oracle outages. ERC165 supportsInterface() exposes both PR_IKPIRewarderClient and PR_IStakingClient.

We built a module that works as a Prediction Bounty. It automatically distributes staking rewards based on real-world outcomes. No team triggers. No manual distribution. An oracle verifies, code executes. 🧵